Closed-Loop Neural Network-Based NMES Control for Human Limb Tracking (original) (raw)
Related papers
Nonlinear tracking control of a human limb via neuromuscular electrical stimulation
2008
Abstract A nonlinear control method is developed in this paper that uses neuromuscular electrical stimulation to control the human quadriceps femoris muscle undergoing non-isometric contractions. The objective of the controller is to position the lower limb of a human along a time-varying trajectory or a desired setpoint. The developed controller does not require a muscle model and can be proven to yield asymptotic stability for a nonlinear muscle model in the presence of bounded nonlinear disturbances.
Nonlinear neuromuscular electrical stimulation tracking control of a human limb
2009
Abstract A high-level objective of neuromuscular electrical stimulation (NMES) is to enable a person to achieve some functional task. Towards this goal, the objective of the current effort is to develop a NMES controller to produce a knee position trajectory that will enable a human shank to track any continuous desired trajectory (or constant setpoint). A nonlinear control method is developed to control the human quadriceps femoris muscle undergoing nonisometric contractions.
Frontiers in Robotics and AI
Neuromuscular electrical stimulation (NMES) is a promising technique to artificially activate muscles as a means to potentially restore the capability to perform functional tasks in persons with neurological disorders. A pervasive problem with NMES is that overstimulation of the muscle (among other factors) leads to rapid muscle fatigue, which limits the use of clinical and commercial NMES systems. The objective of this article is to develop an NMES controller that incorporates the effects of muscle fatigue during NMES-induced non-isometric contraction of the human quadriceps femoris muscle. Our previous work that used the RISE class of non-linear controllers cannot accommodate fatigue and muscle activation dynamics. A totally new control design approach and associated stability proof is required to derive a new class of NMES control design that accounts for muscle fatigue dynamics and a first-order activation dynamics, in addition to the second-order musculoskeletal dynamics. Motivated from a control method for robotic systems in a strict-feedback form, a backstepping based-non-linear NMES controller was designed to accommodate for the additional muscle activation dynamics. Further, experimentally identified estimates of the fatigue and activation dynamics were incorporated in the control design. The developed controller uses a neural networkbased estimate of the musculoskeletal dynamics and error due to fatigue estimation. A globally uniformly ultimately bounded stability is proven the new controller that accounts for an uncertain non-linear muscle model and bounded non-linear disturbances (e.g., spasticity and changing load dynamics). The developed controller was validated through experiments on the left and right legs of 3 able-bodied subjects and was compared with a proportional-derivative (PD) controller and a PD augmented with a neural network. The statistical analysis showed improved control performance compared with the PD controller.
Feedback control methods for task regulation by electrical stimulation of muscles
IEEE Transactions on Biomedical Engineering, 1991
Three feedback control algorithms of varying complexity were compared for controlling three different tasks during electrical stimulation of muscles. Two controllers use stimulus pulse width (or recruitment) modulation to grade muscle force (the fixed parameter, first-order PW controller and the adaptive controller). The third controller varies both stimulus pulse width and period simultaneously for muscle force modulation (the PWlSP controller described in the companion paper). The three tasks tested were isometric torque control, unloaded position tracking, and control of transitions between isometric and unloaded conditions. The first task involved the muscle recruitment nonlinearity. The second task added the effects of muscle length-tension and force-velocity nonlinearities. The third task included a sudden change in external loading conditions. The comparative evaluation was carried out in an intact cat ankle joint with stimulation of tibialis anterior and medial gastrocnemius muscles. The simplest PW controller demonstrated robust control for all tasks. The PW/SP controller improved the performance of the PW controller significantly for control of isometric torque and load transition, but only slightly for control of unloaded joint position. However, the adaptive controller did not consistently achieve a significant improvement in performance compared with the PW controller for any task. Results suggest that muscle length-tension and force-velocity nonlinearities affect the performance of these controllers similarly within the tested ranges of movement amplitudes and speeds. Abrupt changes in the system, such as those due to recruitment nonlinearity and external loading transitions, tend to limit the performance of the adaptive controller. The study provides guidelines for choosing control algorithms for neural prostheses.
Adaptive neural network control of cyclic movements using functional neuromuscular stimulation
IEEE Transactions on Rehabilitation Engineering, 2000
In this study, we evaluated the performance of an adaptive feedforward controller and its ability to automatically develop and customize stimulation patterns for use in functional neuromuscular stimulation (FNS) systems. Results from previous experiments using the pattern generator/pattern shaper (PG/PS) controller to generate isometric contractions demonstrated its ability to adjust stimulation patterns to account for recruitment nonlinearities and muscle dynamics. In this study, the PG/PS controller was tested under isotonic conditions. This evaluation required the PG/PS controller to account for muscle length-tension and force-velocity properties as well as limb dynamics. The performance of the adaptive controller was also compared with that of a proportional-derivative (PD) feedback controller. The PG/PS controller is composed of a neural network system that adaptively filters a periodic signal to produce a muscle stimulation pattern for generating cyclic movements. We used computer-simulated models to determine controller parameters for the PG/PS and PD controller that perform well across a variety of musculoskeletal systems. The controllers were then experimentally evaluated on both legs of two subjects with spinal cord injury. Results indicated that the PG/PS controller was able to achieve and maintain better tracking performance than the PD controller. This study indicates that the PG/PS control system may provide an effective mechanism for automatically customizing stimulation patterns for individuals using FNS systems.
Further Results on Predictor-Based Control of Neuromuscular Electrical Stimulation
Electromechanical delay (EMD) and uncertain non-linear muscle dynamics can cause destabilizing effects and performance loss during closed-loop control of neuromuscular electrical stimulation (NMES). Linear control methods for NMES often perform poorly due to these technical challenges. A new predictor-based closed-loop controller called proportional integral derivative controller with delay compensation (PID-DC) is presented in this paper. The PID-DC controller was designed to compensate for EMDs during NMES. Further, the robust controller can be implemented despite uncertainties or in the absence of model knowledge of the nonlinear musculoskeletal dynamics. Lyapunov stability analysis was used to synthesize the new controller. The effectiveness of the new controller was validated and compared with two recently developed nonlinear NMES controllers, through a series of closed-loop control experiments on four able-bodied human subjects. Experimental results depict statistically significant improved performance with PID-DC. The new controller is shown to be robust to variations in an estimated EMD value. Index Terms-Electromechanical delay, functional electrical stimulation, input delay, Lyapunov-Krasovskii functionals, Lya-punov Methods, neuromuscular electrical stimulation (NMES), nonlinear control.
Functional neurology
This study falls within the ambit of research on functional electrical stimulation for the design of rehabilitation training for spinal cord injured patients. In this context, a crucial issue is the control of the stimulation parameters in order to optimize the patterns of muscle activation and to increase the duration of the exercises. An adaptive control system (NEURADAPT) based on artificial neural networks (ANNs) was developed to control the knee joint in accordance with desired trajectories by stimulating quadriceps muscles. This strategy includes an inverse neural model of the stimulated limb in the feedforward line and a neural network trained on-line in the feedback loop. NEURADAPT was compared with a linear closed-loop proportional integrative derivative (PID) controller and with a model-based neural controller (NEUROPID). Experiments on two subjects (one healthy and one paraplegic) show the good performance of NEURADAPT, which is able to reduce the time lag introduced by t...
IEEE Transactions on Biomedical Engineering, 2009
During the past several years, several strategies have been proposed for control of joint movement in paraplegic subjects using functional electrical stimulation (FES), but developing a control strategy that provides satisfactory tracking performance, to be robust against time-varying properties of muscle-joint dynamics, day-today variations, subject-to-subject variations, muscle fatigue, and external disturbances, and to be easy to apply without any re-identification of plant dynamics during different experiment sessions is still an open problem. In this paper, we propose a novel control methodology that is based on synergistic combination of neural networks with sliding-mode control (SMC) for controlling FES. The main advantage of SMC derives from the property of robustness to system uncertainties and external disturbances. However, the main drawback of the standard sliding modes is mostly related to the so-called chattering caused by the high-frequency control switching. To eliminate the chattering, we couple two neural networks with online learning without any offline training into the SMC. A recurrent neural network is used to model the uncertainties and provide an auxiliary equivalent control to keep the uncertainties to low values, and consequently, to use an SMC with lower switching gain. The second neural network consists of a single neuron and is used as an auxiliary controller. The control law will be switched from the SMC to neural control, when the state trajectory of system enters in some boundary layer around the sliding surface. Extensive simulations and experiments on healthy and paraplegic subjects are provided to demonstrate the robustness, stability, and tracking accuracy of the proposed neuroadaptive SMC. The results show that the neuro-SMC provides accurate tracking control with fast convergence for different reference trajectories and could generate control signals to compensate the muscle fatigue and reject the external disturbance. Index Terms-Functional electrical stimulation (FES), neural network, sliding-mode control (SMC).
Neural Network Controller for an Upper Extremity Neuroprosthesis
Conference Proceedings. 2nd International IEEE EMBS Conference on Neural Engineering, 2005.
The goal of this project is to use functional electrical stimulation (FES) to increase the arm's workspace of individuals with C5/C6 Spinal Cord Injury (SCI); providing more natural movement and thus enhancing functional outcomes. A controller that extracts information from recorded EMG activity of muscles under retained voluntary control and processes these signals to generate the appropriate stimulation levels for paralyzed muscles was designed using a dynamic musculoskeletal model of the arm. Different arm movements were recorded from able bodied subjects and these kinematics served as input to the model. The model was modified to reflect C5/C6 SCI, and inverse simulations were run to provide muscle activation patterns corresponding to the movements recorded. One set of "voluntary" muscles and one set of "stimulated paralyzed" muscles were selected as input and output to the controller based on each muscle's relevance as suggested by the simulations. A neural network controller was trained to predict "stimulated paralyzed" muscle activations using "voluntary" muscle activations as inputs. The neural network controller was able to predict the activation level of three paralyzed muscles with less than 2% error, using four voluntary muscles as inputs.